My team, ThunderBots, had a really good autonomous mode at the Championships, but it was disappointing that we could not get our vision tetra routine to be 100% reliable. We spent a great deal of time on it.
With our practice robot in our workshop, we were able to have the camera find a vision tetra, drive to it, pick it up and cap the center middle goal. We did it several times. But, of course, it needed a little help from hand-held spot lights to illuminate the colored triangles and the disable switch was useful when it took off in the wrong direction....
Much of the time we could get pretty close. Since we had feedback controls for each of the arm movements, the autonomous program we did use was very reliable and easier than the vision tetra version.
But, my personal favorite was in the 2003 Stack Attack game. From the starting position our robot, Yoda, swung a long gravity-assisted telescoping arm that toppled all but the far stack of bins into our side before any robot could make it up the ramp. Many opponents got stuck in the "Sea of Bins". The only way to retract the ginormous arm was to raise it straight up so gravity would pull it back together and a latch would keep it stowed for the rest of the match. Then, still in autonomous, we would drive up the ramp, over the platform, extend 2 wide arms and push all the bins down the ramp.
In 2004, Team 190, Gompei and the HERD, had the most complex and amazing autonomous. The whole robot was designed to make it work.